Although metformin is first line treatment for patients with diabetes (DM), many patients with moderate chronic kidney disease (CKD) do not receive metformin due to concerns about the risk of lactic acidosis. Within the Veterans Health Administration (VHA) 30-40% of incident DM medications are sulfonylureas, rather than metformin, due in part to underlying CKD. Our group and others have demonstrated that among patients with preserved kidney function, metformin use is associated with better cardiovascular, kidney, and survival outcomes compared with sulfonylureas.1-4 The rapidly increasing number of patients with DM and moderate CKD may also derive cardiovascular and renal benefits from metformin, but do not receive this recommended first line therapy due to the Food and Drug Administration (FDA) warning (contraindication at serum creatinine >1.5 mg/dL for men; >1.4mg/dL for women). Guidelines in the United Kingdom, Canada, and Australia emphasize metformin use based on estimated glomerular filtration rate (eGFR) criteria rather than creatinine. Patients with reduced eGFR may use metformin with frequent monitoring and dose reduction, and an eGFR <30 mL/min is considered an absolute contraindication to metformin. Formal petitions to revise the FDA warning to consider eGFR based prescribing have been filed but additional evidence is needed to inform these policy decisions.6, 7 Randomized trials would be ideal, but the large number of patients needed and existing warning make such studies prohibitive. Methodologically-strong comparative effectiveness observational studies are necessary to determine the relative benefits and risks of metformin versus sulfonylureas use among patients with CKD. This proposal will answer the question: Among patients with Stage 3 CKD, what are the absolute risks of lactic acidosis and cardiovascular and renal events for those initially treated with metformin compared to those treated with a sulfonylurea? We will rebuild an inception diabetes cohort utilizing national Veterans Health Administration (VHA) data (corporate data warehouse-vital signs; pharmacy, medical datasets, Medicare/ Medicaid data and vital status files), national death index and data on echocardiograms. Eligible veterans will have a baseline creatinine >1.5 mg/dL or an eGFR between 30 and 59 ml/min and initially treated with metformin (current N=16,891 through FY2008; anticipated rebuilt cohort N~86,924) or sulfonylurea (current N=41,952; anticipated rebuilt cohort N~149,869). The proposed study follows this inception cohort from monotherapy initiation until an outcome or censoring event (leave VHA, change therapy or death). The comparisons of interest will be: metformin vs. sulfonylurea. Persons included are aged >18 years, utilize VHA from 10/1/2000 through 09/30/2014. In this national cohort of veterans with diabetes and impaired renal function: Aim 1 will test the hypothesis that the risk of hospitalization for laboratory-confirmed lactic acidosis in patients treated with metformin is similar to the risk in patients treated with sulfonylureas. Aim 2 will test the hypothesis that fatal or nonfatal cardiovascular events is lower among patients treated with metformin versus sulfonylureas. Aim 3 will test the hypothesis that heart failure hospitalizations or emergency visits is lower among patients treated with metformin versus sulfonylureas. Aim 4 will test the hypothesis that CKD progression and all-cause mortality is lower among patients treated with metformin versus sulfonylureas. Propensity score matching will be used to achieve similar groups. We will compare event rates accounting for baseline and time-varying covariates with marginal structural Cox proportional hazards models (MSM) with ?stabilized? inverse probability treatment weights (IPTW). Our team has been extremely productive over the last 3 years and is poised to conduct a rigorous and necessary evaluation of the effectiveness of DM regimens among patients with CKD. Our publication record attests to our ability to use VHA data for accurate measurement of exposures, outcomes and covariates.